xref: /linux/include/linux/ceph/ceph_fs.h (revision d08008f196107a80c4e88b866d594b88a56ceaa9)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * ceph_fs.h - Ceph constants and data types to share between kernel and
4  * user space.
5  *
6  * Most types in this file are defined as little-endian, and are
7  * primarily intended to describe data structures that pass over the
8  * wire or that are stored on disk.
9  *
10  * LGPL2
11  */
12 
13 #ifndef CEPH_FS_H
14 #define CEPH_FS_H
15 
16 #include <linux/ceph/msgr.h>
17 #include <linux/ceph/rados.h>
18 
19 /*
20  * subprotocol versions.  when specific messages types or high-level
21  * protocols change, bump the affected components.  we keep rev
22  * internal cluster protocols separately from the public,
23  * client-facing protocol.
24  */
25 #define CEPH_OSDC_PROTOCOL   24 /* server/client */
26 #define CEPH_MDSC_PROTOCOL   32 /* server/client */
27 #define CEPH_MONC_PROTOCOL   15 /* server/client */
28 
29 
30 #define CEPH_INO_ROOT   1
31 #define CEPH_INO_CEPH   2            /* hidden .ceph dir */
32 #define CEPH_INO_GLOBAL_SNAPREALM  3 /* global dummy snaprealm */
33 
34 /*
35  * name for "old" CephFS file systems,
36  * see ceph.git e2b151d009640114b2565c901d6f41f6cd5ec652
37  */
38 #define CEPH_OLD_FS_NAME	"cephfs"
39 
40 /* arbitrary limit on max # of monitors (cluster of 3 is typical) */
41 #define CEPH_MAX_MON   31
42 
43 /*
44  * legacy ceph_file_layoute
45  */
46 struct ceph_file_layout_legacy {
47 	/* file -> object mapping */
48 	__le32 fl_stripe_unit;     /* stripe unit, in bytes.  must be multiple
49 				      of page size. */
50 	__le32 fl_stripe_count;    /* over this many objects */
51 	__le32 fl_object_size;     /* until objects are this big, then move to
52 				      new objects */
53 	__le32 fl_cas_hash;        /* UNUSED.  0 = none; 1 = sha256 */
54 
55 	/* pg -> disk layout */
56 	__le32 fl_object_stripe_unit;  /* UNUSED.  for per-object parity, if any */
57 
58 	/* object -> pg layout */
59 	__le32 fl_unused;       /* unused; used to be preferred primary for pg (-1 for none) */
60 	__le32 fl_pg_pool;      /* namespace, crush ruleset, rep level */
61 } __attribute__ ((packed));
62 
63 struct ceph_string;
64 /*
65  * ceph_file_layout - describe data layout for a file/inode
66  */
67 struct ceph_file_layout {
68 	/* file -> object mapping */
69 	u32 stripe_unit;   /* stripe unit, in bytes */
70 	u32 stripe_count;  /* over this many objects */
71 	u32 object_size;   /* until objects are this big */
72 	s64 pool_id;        /* rados pool id */
73 	struct ceph_string __rcu *pool_ns; /* rados pool namespace */
74 };
75 
76 extern int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
77 extern void ceph_file_layout_from_legacy(struct ceph_file_layout *fl,
78 				struct ceph_file_layout_legacy *legacy);
79 extern void ceph_file_layout_to_legacy(struct ceph_file_layout *fl,
80 				struct ceph_file_layout_legacy *legacy);
81 
82 #define CEPH_MIN_STRIPE_UNIT 65536
83 
84 struct ceph_dir_layout {
85 	__u8   dl_dir_hash;   /* see ceph_hash.h for ids */
86 	__u8   dl_unused1;
87 	__u16  dl_unused2;
88 	__u32  dl_unused3;
89 } __attribute__ ((packed));
90 
91 /* crypto algorithms */
92 #define CEPH_CRYPTO_NONE       0x0
93 #define CEPH_CRYPTO_AES        0x1
94 #define CEPH_CRYPTO_AES256KRB5 0x2 /* AES256-CTS-HMAC384-192 */
95 
96 #define CEPH_AES_IV "cephsageyudagreg"
97 
98 /* security/authentication protocols */
99 #define CEPH_AUTH_UNKNOWN	0x0
100 #define CEPH_AUTH_NONE	 	0x1
101 #define CEPH_AUTH_CEPHX	 	0x2
102 
103 #define CEPH_AUTH_MODE_NONE		0
104 #define CEPH_AUTH_MODE_AUTHORIZER	1
105 #define CEPH_AUTH_MODE_MON		10
106 
107 /* msgr2 protocol modes */
108 #define CEPH_CON_MODE_UNKNOWN	0x0
109 #define CEPH_CON_MODE_CRC	0x1
110 #define CEPH_CON_MODE_SECURE	0x2
111 
112 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
113 
114 const char *ceph_auth_proto_name(int proto);
115 const char *ceph_con_mode_name(int mode);
116 
117 /*********************************************
118  * message layer
119  */
120 
121 /*
122  * message types
123  */
124 
125 /* misc */
126 #define CEPH_MSG_SHUTDOWN               1
127 #define CEPH_MSG_PING                   2
128 
129 /* client <-> monitor */
130 #define CEPH_MSG_MON_MAP                4
131 #define CEPH_MSG_MON_GET_MAP            5
132 #define CEPH_MSG_STATFS                 13
133 #define CEPH_MSG_STATFS_REPLY           14
134 #define CEPH_MSG_MON_SUBSCRIBE          15
135 #define CEPH_MSG_MON_SUBSCRIBE_ACK      16
136 #define CEPH_MSG_AUTH			17
137 #define CEPH_MSG_AUTH_REPLY		18
138 #define CEPH_MSG_MON_GET_VERSION        19
139 #define CEPH_MSG_MON_GET_VERSION_REPLY  20
140 
141 /* client <-> mds */
142 #define CEPH_MSG_MDS_MAP                21
143 #define CEPH_MSG_FS_MAP_USER            103
144 
145 #define CEPH_MSG_CLIENT_SESSION         22
146 #define CEPH_MSG_CLIENT_RECONNECT       23
147 
148 #define CEPH_MSG_CLIENT_REQUEST         24
149 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
150 #define CEPH_MSG_CLIENT_REPLY           26
151 #define CEPH_MSG_CLIENT_METRICS         29
152 #define CEPH_MSG_CLIENT_CAPS            0x310
153 #define CEPH_MSG_CLIENT_LEASE           0x311
154 #define CEPH_MSG_CLIENT_SNAP            0x312
155 #define CEPH_MSG_CLIENT_CAPRELEASE      0x313
156 #define CEPH_MSG_CLIENT_QUOTA           0x314
157 
158 /* pool ops */
159 #define CEPH_MSG_POOLOP_REPLY           48
160 #define CEPH_MSG_POOLOP                 49
161 
162 /* mon commands */
163 #define CEPH_MSG_MON_COMMAND            50
164 #define CEPH_MSG_MON_COMMAND_ACK        51
165 
166 /* osd */
167 #define CEPH_MSG_OSD_MAP                41
168 #define CEPH_MSG_OSD_OP                 42
169 #define CEPH_MSG_OSD_OPREPLY            43
170 #define CEPH_MSG_WATCH_NOTIFY           44
171 #define CEPH_MSG_OSD_BACKOFF            61
172 
173 
174 /* watch-notify operations */
175 enum {
176 	CEPH_WATCH_EVENT_NOTIFY		  = 1, /* notifying watcher */
177 	CEPH_WATCH_EVENT_NOTIFY_COMPLETE  = 2, /* notifier notified when done */
178 	CEPH_WATCH_EVENT_DISCONNECT       = 3, /* we were disconnected */
179 };
180 
181 
182 struct ceph_mon_request_header {
183 	__le64 have_version;
184 	__le16 session_mon;
185 	__le64 session_mon_tid;
186 } __attribute__ ((packed));
187 
188 struct ceph_mon_statfs {
189 	struct ceph_mon_request_header monhdr;
190 	struct ceph_fsid fsid;
191 	__u8 contains_data_pool;
192 	__le64 data_pool;
193 } __attribute__ ((packed));
194 
195 struct ceph_statfs {
196 	__le64 kb, kb_used, kb_avail;
197 	__le64 num_objects;
198 } __attribute__ ((packed));
199 
200 struct ceph_mon_statfs_reply {
201 	struct ceph_fsid fsid;
202 	__le64 version;
203 	struct ceph_statfs st;
204 } __attribute__ ((packed));
205 
206 struct ceph_mon_command {
207 	struct ceph_mon_request_header monhdr;
208 	struct ceph_fsid fsid;
209 	__le32 num_strs;         /* always 1 */
210 	__le32 str_len;
211 	char str[];
212 } __attribute__ ((packed));
213 
214 struct ceph_osd_getmap {
215 	struct ceph_mon_request_header monhdr;
216 	struct ceph_fsid fsid;
217 	__le32 start;
218 } __attribute__ ((packed));
219 
220 struct ceph_mds_getmap {
221 	struct ceph_mon_request_header monhdr;
222 	struct ceph_fsid fsid;
223 } __attribute__ ((packed));
224 
225 struct ceph_client_mount {
226 	struct ceph_mon_request_header monhdr;
227 } __attribute__ ((packed));
228 
229 #define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
230 
231 struct ceph_mon_subscribe_item {
232 	__le64 start;
233 	__u8 flags;
234 } __attribute__ ((packed));
235 
236 struct ceph_mon_subscribe_ack {
237 	__le32 duration;         /* seconds */
238 	struct ceph_fsid fsid;
239 } __attribute__ ((packed));
240 
241 #define CEPH_FS_CLUSTER_ID_NONE  -1
242 
243 /*
244  * mdsmap flags
245  */
246 #define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
247 
248 /*
249  * mds states
250  *   > 0 -> in
251  *  <= 0 -> out
252  */
253 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
254 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
255 					  empty log. */
256 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
257 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
258 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
259 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
260 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
261 #define CEPH_MDS_STATE_REPLAYONCE   -9 /* up, replaying an active node's journal */
262 
263 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
264 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
265 					  operations (import, rename, etc.) */
266 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
267 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
268 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
269 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
270 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
271 
272 extern const char *ceph_mds_state_name(int s);
273 
274 
275 /*
276  * metadata lock types.
277  *  - these are bitmasks.. we can compose them
278  *  - they also define the lock ordering by the MDS
279  *  - a few of these are internal to the mds
280  */
281 #define CEPH_LOCK_DVERSION    1
282 #define CEPH_LOCK_DN          2
283 #define CEPH_LOCK_ISNAP       16
284 #define CEPH_LOCK_IVERSION    32    /* mds internal */
285 #define CEPH_LOCK_IFILE       64
286 #define CEPH_LOCK_IAUTH       128
287 #define CEPH_LOCK_ILINK       256
288 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
289 #define CEPH_LOCK_INEST       1024  /* mds internal */
290 #define CEPH_LOCK_IXATTR      2048
291 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
292 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
293 #define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
294 
295 /* client_session ops */
296 enum {
297 	CEPH_SESSION_REQUEST_OPEN,
298 	CEPH_SESSION_OPEN,
299 	CEPH_SESSION_REQUEST_CLOSE,
300 	CEPH_SESSION_CLOSE,
301 	CEPH_SESSION_REQUEST_RENEWCAPS,
302 	CEPH_SESSION_RENEWCAPS,
303 	CEPH_SESSION_STALE,
304 	CEPH_SESSION_RECALL_STATE,
305 	CEPH_SESSION_FLUSHMSG,
306 	CEPH_SESSION_FLUSHMSG_ACK,
307 	CEPH_SESSION_FORCE_RO,
308 	CEPH_SESSION_REJECT,
309 	CEPH_SESSION_REQUEST_FLUSH_MDLOG,
310 };
311 
312 #define CEPH_SESSION_BLOCKLISTED	(1 << 0)  /* session blocklisted */
313 
314 extern const char *ceph_session_op_name(int op);
315 
316 struct ceph_mds_session_head {
317 	__le32 op;
318 	__le64 seq;
319 	struct ceph_timespec stamp;
320 	__le32 max_caps, max_leases;
321 } __attribute__ ((packed));
322 
323 /* client_request */
324 /*
325  * metadata ops.
326  *  & 0x001000 -> write op
327  *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
328  &  & 0x100000 -> use weird ino/path trace
329  */
330 #define CEPH_MDS_OP_WRITE        0x001000
331 enum {
332 	CEPH_MDS_OP_LOOKUP     = 0x00100,
333 	CEPH_MDS_OP_GETATTR    = 0x00101,
334 	CEPH_MDS_OP_LOOKUPHASH = 0x00102,
335 	CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
336 	CEPH_MDS_OP_LOOKUPINO  = 0x00104,
337 	CEPH_MDS_OP_LOOKUPNAME = 0x00105,
338 	CEPH_MDS_OP_GETVXATTR  = 0x00106,
339 
340 	CEPH_MDS_OP_SETXATTR   = 0x01105,
341 	CEPH_MDS_OP_RMXATTR    = 0x01106,
342 	CEPH_MDS_OP_SETLAYOUT  = 0x01107,
343 	CEPH_MDS_OP_SETATTR    = 0x01108,
344 	CEPH_MDS_OP_SETFILELOCK= 0x01109,
345 	CEPH_MDS_OP_GETFILELOCK= 0x00110,
346 	CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
347 
348 	CEPH_MDS_OP_MKNOD      = 0x01201,
349 	CEPH_MDS_OP_LINK       = 0x01202,
350 	CEPH_MDS_OP_UNLINK     = 0x01203,
351 	CEPH_MDS_OP_RENAME     = 0x01204,
352 	CEPH_MDS_OP_MKDIR      = 0x01220,
353 	CEPH_MDS_OP_RMDIR      = 0x01221,
354 	CEPH_MDS_OP_SYMLINK    = 0x01222,
355 
356 	CEPH_MDS_OP_CREATE     = 0x01301,
357 	CEPH_MDS_OP_OPEN       = 0x00302,
358 	CEPH_MDS_OP_READDIR    = 0x00305,
359 
360 	CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
361 	CEPH_MDS_OP_MKSNAP     = 0x01400,
362 	CEPH_MDS_OP_RMSNAP     = 0x01401,
363 	CEPH_MDS_OP_LSSNAP     = 0x00402,
364 	CEPH_MDS_OP_RENAMESNAP = 0x01403,
365 };
366 
367 #define IS_CEPH_MDS_OP_NEWINODE(op) (op == CEPH_MDS_OP_CREATE     || \
368 				     op == CEPH_MDS_OP_MKNOD      || \
369 				     op == CEPH_MDS_OP_MKDIR      || \
370 				     op == CEPH_MDS_OP_SYMLINK)
371 
372 extern const char *ceph_mds_op_name(int op);
373 
374 #define CEPH_SETATTR_MODE              (1 << 0)
375 #define CEPH_SETATTR_UID               (1 << 1)
376 #define CEPH_SETATTR_GID               (1 << 2)
377 #define CEPH_SETATTR_MTIME             (1 << 3)
378 #define CEPH_SETATTR_ATIME             (1 << 4)
379 #define CEPH_SETATTR_SIZE              (1 << 5)
380 #define CEPH_SETATTR_CTIME             (1 << 6)
381 #define CEPH_SETATTR_MTIME_NOW         (1 << 7)
382 #define CEPH_SETATTR_ATIME_NOW         (1 << 8)
383 #define CEPH_SETATTR_BTIME             (1 << 9)
384 #define CEPH_SETATTR_KILL_SGUID        (1 << 10)
385 #define CEPH_SETATTR_FSCRYPT_AUTH      (1 << 11)
386 #define CEPH_SETATTR_FSCRYPT_FILE      (1 << 12)
387 
388 /*
389  * Ceph setxattr request flags.
390  */
391 #define CEPH_XATTR_CREATE  (1 << 0)
392 #define CEPH_XATTR_REPLACE (1 << 1)
393 #define CEPH_XATTR_REMOVE  (1 << 31)
394 
395 /*
396  * readdir request flags;
397  */
398 #define CEPH_READDIR_REPLY_BITFLAGS	(1<<0)
399 
400 /*
401  * readdir reply flags.
402  */
403 #define CEPH_READDIR_FRAG_END		(1<<0)
404 #define CEPH_READDIR_FRAG_COMPLETE	(1<<8)
405 #define CEPH_READDIR_HASH_ORDER		(1<<9)
406 #define CEPH_READDIR_OFFSET_HASH	(1<<10)
407 
408 /*
409  * open request flags
410  */
411 #define CEPH_O_RDONLY		00000000
412 #define CEPH_O_WRONLY		00000001
413 #define CEPH_O_RDWR		00000002
414 #define CEPH_O_CREAT		00000100
415 #define CEPH_O_EXCL		00000200
416 #define CEPH_O_TRUNC		00001000
417 #define CEPH_O_DIRECTORY	00200000
418 #define CEPH_O_NOFOLLOW		00400000
419 
420 union ceph_mds_request_args {
421 	struct {
422 		__le32 mask;                 /* CEPH_CAP_* */
423 	} __attribute__ ((packed)) getattr;
424 	struct {
425 		__le32 mode;
426 		__le32 uid;
427 		__le32 gid;
428 		struct ceph_timespec mtime;
429 		struct ceph_timespec atime;
430 		__le64 size, old_size;       /* old_size needed by truncate */
431 		__le32 mask;                 /* CEPH_SETATTR_* */
432 	} __attribute__ ((packed)) setattr;
433 	struct {
434 		__le32 frag;                 /* which dir fragment */
435 		__le32 max_entries;          /* how many dentries to grab */
436 		__le32 max_bytes;
437 		__le16 flags;
438 		__le32 offset_hash;
439 	} __attribute__ ((packed)) readdir;
440 	struct {
441 		__le32 mode;
442 		__le32 rdev;
443 	} __attribute__ ((packed)) mknod;
444 	struct {
445 		__le32 mode;
446 	} __attribute__ ((packed)) mkdir;
447 	struct {
448 		__le32 flags;
449 		__le32 mode;
450 		__le32 stripe_unit;          /* layout for newly created file */
451 		__le32 stripe_count;         /* ... */
452 		__le32 object_size;
453 		__le32 pool;
454 		__le32 mask;                 /* CEPH_CAP_* */
455 		__le64 old_size;
456 	} __attribute__ ((packed)) open;
457 	struct {
458 		__le32 flags;
459 		__le32 osdmap_epoch; /* used for setting file/dir layouts */
460 	} __attribute__ ((packed)) setxattr;
461 	struct {
462 		struct ceph_file_layout_legacy layout;
463 	} __attribute__ ((packed)) setlayout;
464 	struct {
465 		__u8 rule; /* currently fcntl or flock */
466 		__u8 type; /* shared, exclusive, remove*/
467 		__le64 owner; /* owner of the lock */
468 		__le64 pid; /* process id requesting the lock */
469 		__le64 start; /* initial location to lock */
470 		__le64 length; /* num bytes to lock from start */
471 		__u8 wait; /* will caller wait for lock to become available? */
472 	} __attribute__ ((packed)) filelock_change;
473 	struct {
474 		__le32 mask;                 /* CEPH_CAP_* */
475 		__le64 snapid;
476 		__le64 parent;
477 		__le32 hash;
478 	} __attribute__ ((packed)) lookupino;
479 } __attribute__ ((packed));
480 
481 union ceph_mds_request_args_ext {
482 	union ceph_mds_request_args old;
483 	struct {
484 		__le32 mode;
485 		__le32 uid;
486 		__le32 gid;
487 		struct ceph_timespec mtime;
488 		struct ceph_timespec atime;
489 		__le64 size, old_size;       /* old_size needed by truncate */
490 		__le32 mask;                 /* CEPH_SETATTR_* */
491 		struct ceph_timespec btime;
492 	} __attribute__ ((packed)) setattr_ext;
493 };
494 
495 #define CEPH_MDS_FLAG_REPLAY		1 /* this is a replayed op */
496 #define CEPH_MDS_FLAG_WANT_DENTRY	2 /* want dentry in reply */
497 #define CEPH_MDS_FLAG_ASYNC		4 /* request is asynchronous */
498 
499 struct ceph_mds_request_head_legacy {
500 	__le64 oldest_client_tid;
501 	__le32 mdsmap_epoch;           /* on client */
502 	__le32 flags;                  /* CEPH_MDS_FLAG_* */
503 	__u8 num_retry, num_fwd;       /* count retry, fwd attempts */
504 	__le16 num_releases;           /* # include cap/lease release records */
505 	__le32 op;                     /* mds op code */
506 	__le32 caller_uid, caller_gid;
507 	__le64 ino;                    /* use this ino for openc, mkdir, mknod,
508 					  etc. (if replaying) */
509 	union ceph_mds_request_args args;
510 } __attribute__ ((packed));
511 
512 #define CEPH_MDS_REQUEST_HEAD_VERSION  3
513 
514 struct ceph_mds_request_head {
515 	__le16 version;                /* struct version */
516 	__le64 oldest_client_tid;
517 	__le32 mdsmap_epoch;           /* on client */
518 	__le32 flags;                  /* CEPH_MDS_FLAG_* */
519 	__u8 num_retry, num_fwd;       /* legacy count retry and fwd attempts */
520 	__le16 num_releases;           /* # include cap/lease release records */
521 	__le32 op;                     /* mds op code */
522 	__le32 caller_uid, caller_gid;
523 	__le64 ino;                    /* use this ino for openc, mkdir, mknod,
524 					  etc. (if replaying) */
525 	union ceph_mds_request_args_ext args;
526 
527 	__le32 ext_num_retry;          /* new count retry attempts */
528 	__le32 ext_num_fwd;            /* new count fwd attempts */
529 
530 	__le32 struct_len;             /* to store size of struct ceph_mds_request_head */
531 	__le32 owner_uid, owner_gid;   /* used for OPs which create inodes */
532 } __attribute__ ((packed));
533 
534 /* cap/lease release record */
535 struct ceph_mds_request_release {
536 	__le64 ino, cap_id;            /* ino and unique cap id */
537 	__le32 caps, wanted;           /* new issued, wanted */
538 	__le32 seq, issue_seq, mseq;
539 	__le32 dname_seq;              /* if releasing a dentry lease, a */
540 	__le32 dname_len;              /* string follows. */
541 } __attribute__ ((packed));
542 
543 /* client reply */
544 struct ceph_mds_reply_head {
545 	__le32 op;
546 	__le32 result;
547 	__le32 mdsmap_epoch;
548 	__u8 safe;                     /* true if committed to disk */
549 	__u8 is_dentry, is_target;     /* true if dentry, target inode records
550 					  are included with reply */
551 } __attribute__ ((packed));
552 
553 /* one for each node split */
554 struct ceph_frag_tree_split {
555 	__le32 frag;                   /* this frag splits... */
556 	__le32 by;                     /* ...by this many bits */
557 } __attribute__ ((packed));
558 
559 struct ceph_frag_tree_head {
560 	__le32 nsplits;                /* num ceph_frag_tree_split records */
561 	struct ceph_frag_tree_split splits[];
562 } __attribute__ ((packed));
563 
564 /* capability issue, for bundling with mds reply */
565 struct ceph_mds_reply_cap {
566 	__le32 caps, wanted;           /* caps issued, wanted */
567 	__le64 cap_id;
568 	__le32 seq, mseq;
569 	__le64 realm;                  /* snap realm */
570 	__u8 flags;                    /* CEPH_CAP_FLAG_* */
571 } __attribute__ ((packed));
572 
573 #define CEPH_CAP_FLAG_AUTH	(1 << 0)  /* cap is issued by auth mds */
574 #define CEPH_CAP_FLAG_RELEASE	(1 << 1)  /* release the cap */
575 
576 /* inode record, for bundling with mds reply */
577 struct ceph_mds_reply_inode {
578 	__le64 ino;
579 	__le64 snapid;
580 	__le32 rdev;
581 	__le64 version;                /* inode version */
582 	__le64 xattr_version;          /* version for xattr blob */
583 	struct ceph_mds_reply_cap cap; /* caps issued for this inode */
584 	struct ceph_file_layout_legacy layout;
585 	struct ceph_timespec ctime, mtime, atime;
586 	__le32 time_warp_seq;
587 	__le64 size, max_size, truncate_size;
588 	__le32 truncate_seq;
589 	__le32 mode, uid, gid;
590 	__le32 nlink;
591 	__le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
592 	struct ceph_timespec rctime;
593 	struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
594 } __attribute__ ((packed));
595 /* followed by frag array, symlink string, dir layout, xattr blob */
596 
597 /* reply_lease follows dname, and reply_inode */
598 struct ceph_mds_reply_lease {
599 	__le16 mask;            /* lease type(s) */
600 	__le32 duration_ms;     /* lease duration */
601 	__le32 seq;
602 } __attribute__ ((packed));
603 
604 #define CEPH_LEASE_VALID        (1 | 2) /* old and new bit values */
605 #define CEPH_LEASE_PRIMARY_LINK 4       /* primary linkage */
606 
607 struct ceph_mds_reply_dirfrag {
608 	__le32 frag;            /* fragment */
609 	__le32 auth;            /* auth mds, if this is a delegation point */
610 	__le32 ndist;           /* number of mds' this is replicated on */
611 	__le32 dist[];
612 } __attribute__ ((packed));
613 
614 #define CEPH_LOCK_FCNTL		1
615 #define CEPH_LOCK_FLOCK		2
616 #define CEPH_LOCK_FCNTL_INTR    3
617 #define CEPH_LOCK_FLOCK_INTR    4
618 
619 
620 #define CEPH_LOCK_SHARED   1
621 #define CEPH_LOCK_EXCL     2
622 #define CEPH_LOCK_UNLOCK   4
623 
624 struct ceph_filelock {
625 	__le64 start;/* file offset to start lock at */
626 	__le64 length; /* num bytes to lock; 0 for all following start */
627 	__le64 client; /* which client holds the lock */
628 	__le64 owner; /* owner the lock */
629 	__le64 pid; /* process id holding the lock on the client */
630 	__u8 type; /* shared lock, exclusive lock, or unlock */
631 } __attribute__ ((packed));
632 
633 
634 /* file access modes */
635 #define CEPH_FILE_MODE_PIN        0
636 #define CEPH_FILE_MODE_RD         1
637 #define CEPH_FILE_MODE_WR         2
638 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
639 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
640 #define CEPH_FILE_MODE_BITS       4
641 #define CEPH_FILE_MODE_MASK       ((1 << CEPH_FILE_MODE_BITS) - 1)
642 
643 int ceph_flags_to_mode(int flags);
644 
645 #define CEPH_INLINE_NONE	((__u64)-1)
646 
647 /* capability bits */
648 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
649 
650 /* generic cap bits */
651 #define CEPH_CAP_GSHARED     1  /* client can reads */
652 #define CEPH_CAP_GEXCL       2  /* client can read and update */
653 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
654 #define CEPH_CAP_GRD         8  /* (file) client can read */
655 #define CEPH_CAP_GWR        16  /* (file) client can write */
656 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
657 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
658 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
659 
660 #define CEPH_CAP_SIMPLE_BITS  2
661 #define CEPH_CAP_FILE_BITS    8
662 
663 /* per-lock shift */
664 #define CEPH_CAP_SAUTH      2
665 #define CEPH_CAP_SLINK      4
666 #define CEPH_CAP_SXATTR     6
667 #define CEPH_CAP_SFILE      8
668 #define CEPH_CAP_SFLOCK    20
669 
670 #define CEPH_CAP_BITS      22
671 
672 /* composed values */
673 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
674 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
675 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
676 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
677 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
678 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
679 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
680 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
681 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
682 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
683 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
684 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
685 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
686 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
687 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
688 #define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
689 #define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
690 
691 
692 /* cap masks (for getattr) */
693 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
694 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
695 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
696 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
697 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
698 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
699 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
700 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
701 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
702 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
703 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
704 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
705 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |			\
706 				 CEPH_CAP_AUTH_SHARED |	\
707 				 CEPH_CAP_LINK_SHARED |	\
708 				 CEPH_CAP_FILE_SHARED |	\
709 				 CEPH_CAP_XATTR_SHARED)
710 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \
711 				   CEPH_CAP_FILE_RD)
712 #define CEPH_STAT_RSTAT CEPH_CAP_FILE_WREXTEND
713 
714 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |			\
715 			      CEPH_CAP_LINK_SHARED |			\
716 			      CEPH_CAP_XATTR_SHARED |			\
717 			      CEPH_CAP_FILE_SHARED)
718 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |	\
719 			   CEPH_CAP_FILE_CACHE)
720 
721 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |		\
722 			   CEPH_CAP_LINK_EXCL |		\
723 			   CEPH_CAP_XATTR_EXCL |	\
724 			   CEPH_CAP_FILE_EXCL)
725 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \
726 			      CEPH_CAP_FILE_SHARED)
727 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER |	\
728 			      CEPH_CAP_FILE_EXCL)
729 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
730 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
731 			   CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
732 			   CEPH_CAP_PIN)
733 #define CEPH_CAP_ALL_FILE (CEPH_CAP_PIN | CEPH_CAP_ANY_SHARED | \
734 			   CEPH_CAP_AUTH_EXCL | CEPH_CAP_XATTR_EXCL | \
735 			   CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR)
736 
737 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
738 			CEPH_LOCK_IXATTR)
739 
740 /* cap masks async dir operations */
741 #define CEPH_CAP_DIR_CREATE	CEPH_CAP_FILE_CACHE
742 #define CEPH_CAP_DIR_UNLINK	CEPH_CAP_FILE_RD
743 #define CEPH_CAP_ANY_DIR_OPS	(CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_RD | \
744 				 CEPH_CAP_FILE_WREXTEND | CEPH_CAP_FILE_LAZYIO)
745 
746 int ceph_caps_for_mode(int mode);
747 
748 enum {
749 	CEPH_CAP_OP_GRANT,         /* mds->client grant */
750 	CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
751 	CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
752 	CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
753 	CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
754 	CEPH_CAP_OP_UPDATE,        /* client->mds update */
755 	CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
756 	CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
757 	CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
758 	CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
759 	CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
760 	CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
761 	CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
762 };
763 
764 extern const char *ceph_cap_op_name(int op);
765 
766 /* flags field in client cap messages (version >= 10) */
767 #define CEPH_CLIENT_CAPS_SYNC			(1<<0)
768 #define CEPH_CLIENT_CAPS_NO_CAPSNAP		(1<<1)
769 #define CEPH_CLIENT_CAPS_PENDING_CAPSNAP	(1<<2)
770 
771 /*
772  * caps message, used for capability callbacks, acks, requests, etc.
773  */
774 struct ceph_mds_caps {
775 	__le32 op;                  /* CEPH_CAP_OP_* */
776 	__le64 ino, realm;
777 	__le64 cap_id;
778 	__le32 seq, issue_seq;
779 	__le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
780 	__le32 migrate_seq;
781 	__le64 snap_follows;
782 	__le32 snap_trace_len;
783 
784 	/* authlock */
785 	__le32 uid, gid, mode;
786 
787 	/* linklock */
788 	__le32 nlink;
789 
790 	/* xattrlock */
791 	__le32 xattr_len;
792 	__le64 xattr_version;
793 
794 	/* a union of non-export and export bodies. */
795 	__le64 size, max_size, truncate_size;
796 	__le32 truncate_seq;
797 	struct ceph_timespec mtime, atime, ctime;
798 	struct ceph_file_layout_legacy layout;
799 	__le32 time_warp_seq;
800 } __attribute__ ((packed));
801 
802 struct ceph_mds_cap_peer {
803 	__le64 cap_id;
804 	__le32 issue_seq;
805 	__le32 mseq;
806 	__le32 mds;
807 	__u8   flags;
808 } __attribute__ ((packed));
809 
810 /* cap release msg head */
811 struct ceph_mds_cap_release {
812 	__le32 num;                /* number of cap_items that follow */
813 } __attribute__ ((packed));
814 
815 struct ceph_mds_cap_item {
816 	__le64 ino;
817 	__le64 cap_id;
818 	__le32 migrate_seq, issue_seq;
819 } __attribute__ ((packed));
820 
821 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
822 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
823 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
824 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
825 
826 extern const char *ceph_lease_op_name(int o);
827 
828 /* lease msg header */
829 struct ceph_mds_lease {
830 	__u8 action;            /* CEPH_MDS_LEASE_* */
831 	__le16 mask;            /* which lease */
832 	__le64 ino;
833 	__le64 first, last;     /* snap range */
834 	__le32 seq;
835 	__le32 duration_ms;     /* duration of renewal */
836 } __attribute__ ((packed));
837 /* followed by a __le32+string for dname */
838 
839 /* client reconnect */
840 struct ceph_mds_cap_reconnect {
841 	__le64 cap_id;
842 	__le32 wanted;
843 	__le32 issued;
844 	__le64 snaprealm;
845 	__le64 pathbase;        /* base ino for our path to this ino */
846 	__le32 flock_len;       /* size of flock state blob, if any */
847 } __attribute__ ((packed));
848 /* followed by flock blob */
849 
850 struct ceph_mds_cap_reconnect_v1 {
851 	__le64 cap_id;
852 	__le32 wanted;
853 	__le32 issued;
854 	__le64 size;
855 	struct ceph_timespec mtime, atime;
856 	__le64 snaprealm;
857 	__le64 pathbase;        /* base ino for our path to this ino */
858 } __attribute__ ((packed));
859 
860 struct ceph_mds_snaprealm_reconnect {
861 	__le64 ino;     /* snap realm base */
862 	__le64 seq;     /* snap seq for this snap realm */
863 	__le64 parent;  /* parent realm */
864 } __attribute__ ((packed));
865 
866 /*
867  * snaps
868  */
869 enum {
870 	CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
871 	CEPH_SNAP_OP_CREATE,
872 	CEPH_SNAP_OP_DESTROY,
873 	CEPH_SNAP_OP_SPLIT,
874 };
875 
876 extern const char *ceph_snap_op_name(int o);
877 
878 /* snap msg header */
879 struct ceph_mds_snap_head {
880 	__le32 op;                /* CEPH_SNAP_OP_* */
881 	__le64 split;             /* ino to split off, if any */
882 	__le32 num_split_inos;    /* # inos belonging to new child realm */
883 	__le32 num_split_realms;  /* # child realms udner new child realm */
884 	__le32 trace_len;         /* size of snap trace blob */
885 } __attribute__ ((packed));
886 /* followed by split ino list, then split realms, then the trace blob */
887 
888 /*
889  * encode info about a snaprealm, as viewed by a client
890  */
891 struct ceph_mds_snap_realm {
892 	__le64 ino;           /* ino */
893 	__le64 created;       /* snap: when created */
894 	__le64 parent;        /* ino: parent realm */
895 	__le64 parent_since;  /* snap: same parent since */
896 	__le64 seq;           /* snap: version */
897 	__le32 num_snaps;
898 	__le32 num_prior_parent_snaps;
899 } __attribute__ ((packed));
900 /* followed by my snap list, then prior parent snap list */
901 
902 /*
903  * quotas
904  */
905 struct ceph_mds_quota {
906 	__le64 ino;		/* ino */
907 	struct ceph_timespec rctime;
908 	__le64 rbytes;		/* dir stats */
909 	__le64 rfiles;
910 	__le64 rsubdirs;
911 	__u8 struct_v;		/* compat */
912 	__u8 struct_compat;
913 	__le32 struct_len;
914 	__le64 max_bytes;	/* quota max. bytes */
915 	__le64 max_files;	/* quota max. files */
916 } __attribute__ ((packed));
917 
918 #endif
919